Tips: 快速查询定位最慢的SQL

最耗时SQL:

SELECT *
   FROM (SELECT C.SQL_ID,
                (C.ETIME - NVL(O.ELAPSED_TIME_TOTAL, 0)) / 1000000 ELAPSED_DELTA_TIME
           FROM (SELECT *
                   FROM (SELECT SQL_ID,
                                PLAN_HASH_VALUE,
                                SUM(ELAPSED_TIME) ETIME
                           FROM V$SQL S
                          WHERE S.PLAN_HASH_VALUE <> 0
                          GROUP BY SQL_ID, PLAN_HASH_VALUE
                         HAVING MAX(LAST_ACTIVE_TIME) > SYSDATE - 1 / 24
                         ORDER BY SUM(ELAPSED_TIME) DESC)
                  WHERE ROWNUM < 100) C,
                (SELECT SQL_ID, ELAPSED_TIME_TOTAL, PLAN_HASH_VALUE
                   FROM DBA_HIST_SQLSTAT HIS
                  WHERE HIS.SNAP_ID =
                        (SELECT MAX(SNAP_ID) FROM DBA_HIST_SNAPSHOT)
                    AND HIS.INSTANCE_NUMBER = USERENV('instance')) O
          WHERE C.SQL_ID = O.SQL_ID
            AND C.PLAN_HASH_VALUE = O.PLAN_HASH_VALUE
          ORDER BY ELAPSED_DELTA_TIME DESC)
 WHERE ROWNUM < 10;

会得到类似下面的结果：

SQL_ID                                  ELAPSED_DELTA_TIME
--------------------------------------- ------------------
42bd8gf0xd720                                   2765.146325
fr49k0t6qdz6c                                      .910986
61xywzr81rbm4                                      .459234
1yd0xckwg8mrq                                      .295755
ag7apfq5716hc                                      .270697
f9rzyzsypmnkm                                      .267293
cr777xwby6a5d                                      .174171
5y6xxf087crn5                                         .168
b6km56yym0gtv                                      .152382

2 . 得到SQL_ID “42bd8gf0xd720”, 根据SQLID 拉取执行计划，即可根据执行计划分析:

select * from table(dbms_xplan.display_awr('42bd8gf0xd720'));

3 . 如果知道sql语句，也可以通过explan plan for+语句 拉取执行计划
如果要添加索引优化，记得加 online

CREATE INDEX "USERNAME"."IDX_索引名" ON "拥有者"."表名" ("列名") online;

4 .调优后如果添加了index或者做了改动，要重新收集下表的统计信息：

exec dbms_stats.gather_table_stats(ownname => '表拥有着',tabname => '表名' ,method_opt => 'for all indexed columns' ,cascade => true);